Tensorflow.js tf.io.http() Function

Tensorflow.js is an open-source library that is developed by Google for running machine learning models as well as deep learning neural networks in the browser or node environment.

The .io.http() function is used to generate an IOHandler subset which transmits model artifacts to the HTTP server. Moreover, an HTTP request of the _multi-part or _form-data mime type shall be transmitted to the path URL. Where, the form-data contains artifacts which describe the model topology and/or weights of the model.

Syntax:

tf.io.http(path, loadOptions?)

Parameters:

path: The stated URL path to the model. Moreover, it could be a complete HTTP path i.e. ‘http://localhost:8000/model-upload’ or a comparable path i.e. ‘./model-upload’. It is of type string.
loadOptions: The stated configuration for the purpose of loading. It is optional and is of type LoadOptions. It comprises the following fields:
1. weightPathPrefix: The stated optional prefix of path for the weight files. In addition, the by default value of this is evaluated from the path param.
2. fetchFunc: The stated optional custom _fetchfunction.
3. onProgress: The stated optional progress callback function, it is released at regular intervals prior to the completion of load.

Return Value: It returns IOHandler.

Example 1:

Javascript

// Importing the tensorflow.js library 

import * as tf from "@tensorflow/tfjs"

// Calling io.http() method 

const result = tf.io.http('https://js.tensorflow.org/api/latest/#io.http'); 

// Printing output 
console.log(result);

Output:

{
  "DEFAULT_METHOD": "POST",
  "path": "https://js.tensorflow.org/api/latest/#io.http",
  "requestInit": {}
}

Example 2:

Javascript

// Importing the tensorflow.js library 

import * as tf from "@tensorflow/tfjs"

// Creating model 
const model = tf.sequential(); 

// Adding layer to the model 
model.add( 

     tf.layers.dense({units: 2, inputShape: [10]})); 

// Calling io.http() method within  
// save() method 
const result = await model.save(tf.io.http( 

     'https://js.tensorflow.org/api/latest/#io.http')); 

// Printing output 
console.log(result);

Output:

{
  "modelArtifactsInfo": {
    "dateSaved": "2021-08-26T08:43:49.648Z",
    "modelTopologyType": "JSON",
    "modelTopologyBytes": 611,
    "weightSpecsBytes": 124,
    "weightDataBytes": 88
  },
  "responses": [
    {}
  ]
}

Reference: https://js.tensorflow.org/api/latest/#io.http

Article Tags :

JavaScript

Web Technologies

Tensorflow.js